ICBCCR: A new path for harmonization and clinical verification on a multi-center real-world cohort of 278703.

Abstract

Background: In this study, we aimed to develop a novel harmonization algorithm to achieve the conversion from current results to harmonized results directly and perform a clinical verification on a multi-center real-world cohort of 278,703.

Methods: Sixty patient samples were collected as harmonization reference materials to develop and internally validate the harmonization algorithm. We assessed the harmonization effectiveness through multiple dimensions such as method comparison and evaluation. Then, the harmonization algorithm was performed on 278,703 sample results. We compared the distributions of the harmonized results from different in vitro diagnostic measurement devices through the Bhattacharyya distance.

Results: After harmonized by the novel algorithm, equivalent results were obtained among 5 in vitro diagnostic measurement devices for MYO with mean percent differences from -1.5 % to -0.1 % and the 95 % limits of agreement from 14.1 % to 20.3 % and for CK-MB with mean percent differences from -0.2 % to 1.1 % and the 95 % limits of agreement from 14.5 % to 36.6 %. The coefficients of variation of the results for the same sample presented a significant decrease from 20.6 % to 4.6 % for MYO and from 38.6 % to 5.3 % for CK-MB. Furthermore, after the harmonization algorithm was performed on 278,703 sample results from the real world, the Bhattacharyya distance decreased from 0.0046 to 0.0019 for MYO and from 0.0949 to 0.0049 for CK-MB after harmonization, respectively.

Conclusions: The novel algorithm is a new path for harmonization only based on current results and its clinical applicability has been fully verified in the real world.